Adaptive LBP Features to Describe Image for Detecting Variant Patterns in Face Texture for Age Estimation
Face texture has different patterns of changes, which can provide different types of information regarding inter-class and intra-class changes of different classification problems. This paper proposes a new face descriptor to provide a set of features that represent variant patterns of image texture. Modifying standard LBP descriptor provides features that detect variant patterns considering changes in all neighbour pixels. Proposed technique produces binary code by comparing each neighbouring pixel with the average of mask pixels rather than the central pixel of the mask. While the standard operator of LBP provided encouraging results in identifying problems like face and character recognition, proposed face descriptor serves different problems such as age estimation and gender recognition. Proposed face descriptor provides features with more robustness to classification challenges like illumination.